Finding line segments in range images

摘要

The problem of fitting straight lines to a number of data points arises frequently in image processing and computer vision. In this paper, we consider two approaches to solve the line detection problem in range images. The first is a simple 3D approach that finds the best line among a randomly selected subset of lines each through a pair of edge points. The second approach, based on tabu search, optimizes the fit of the lines to the points. We present experimental results obtained with the two approaches on range images of polyhedral objects.